[BUG] softlockup observed with 64k block size on ppc64

[BUG] softlockup observed with 64k block size on ppc64

[Chandan Rajendra]

Hi,

The following soft lockup occurs when fstests' ext4/303 test is executed with
64k block size on a ppc64 machine running linux-4.8 kernel.

NMI watchdog: BUG: soft lockup - CPU#1 stuck for 23s! [fio:12268]
Modules linked in:
CPU: 1 PID: 12268 Comm: fio Not tainted 4.8.0 #1
task: c00000062d344900 task.stack: c0000006283c4000
NIP: c0000000003a55e4 LR: c0000000003a5564 CTR: 0000000000000000
REGS: c0000006283c6fb0 TRAP: 0901   Not tainted  (4.8.0)
MSR: 8000000002009032 <SF,VEC,EE,ME,IR,DR,RI>  CR: 44428848  XER: 00000000
CFAR: c0000000003a5670 SOFTE: 1
GPR00: c0000000003a79cc c0000006283c7230 c00000000110f100 0000000000008010
GPR04: 0000000000000000 c000000630fd7000 fffffffffffffffe 0000000000000010
GPR08: 0000000000008010 ffffffffffffffff 000000000000000f 0000000081860c2d
GPR12: 0000000024428244 c00000000fe00400
NIP [c0000000003a55e4] .ext4_mb_generate_buddy+0x1a4/0x450
LR [c0000000003a5564] .ext4_mb_generate_buddy+0x124/0x450
Call Trace:
[c0000006283c7230] [0000000000000002] 0x2 (unreliable)
[c0000006283c7340] [c0000000003a79cc] .ext4_mb_init_cache+0x43c/0x830
[c0000006283c7470] [c0000000003a8040] .ext4_mb_init_group+0x280/0x350
[c0000006283c7540] [c0000000003a8340] .ext4_mb_good_group+0x230/0x240
[c0000006283c75e0] [c0000000003abf8c] .ext4_mb_regular_allocator+0x2ec/0x540
[c0000006283c7720] [c0000000003ae1c0] .ext4_mb_new_blocks+0x680/0xd80
[c0000006283c7820] [c00000000039ee98] .ext4_ext_map_blocks+0x758/0x1d20
[c0000006283c79d0] [c000000000365e24] .ext4_map_blocks+0x1c4/0x750
[c0000006283c7ab0] [c000000000399230] .ext4_alloc_file_blocks.isra.6+0x170/0x3e0
[c0000006283c7be0] [c0000000003a1e24] .ext4_fallocate+0x2f4/0xa20
[c0000006283c7cf0] [c000000000283f54] .vfs_fallocate+0x144/0x290
[c0000006283c7d90] [c000000000285298] .SyS_fallocate+0x58/0xd0
[c0000006283c7e30] [c000000000009560] system_call+0x38/0x108
Instruction dump:
889e0014 7f042030 5484043e 7e892378 7d2907b4 7d4900d0 7d494838 7d070034
7d290074 2147001f 2129003f 7f895000 <409d0008> 7d495378 7d2a07b4 2e890000

-- 
chandan

Re: [BUG] softlockup observed with 64k block size on ppc64

[Jan Kara]

Hi!

On Mon 24-10-16 14:10:02, Chandan Rajendra wrote:
> The following soft lockup occurs when fstests' ext4/303 test is executed with
> 64k block size on a ppc64 machine running linux-4.8 kernel.

Thanks for report! I had a look into the code and I'm not sure why this
happens. Is this reproducible? Did it happen also with older kernels?

								Honza
> 
> NMI watchdog: BUG: soft lockup - CPU#1 stuck for 23s! [fio:12268]
> Modules linked in:
> CPU: 1 PID: 12268 Comm: fio Not tainted 4.8.0 #1
> task: c00000062d344900 task.stack: c0000006283c4000
> NIP: c0000000003a55e4 LR: c0000000003a5564 CTR: 0000000000000000
> REGS: c0000006283c6fb0 TRAP: 0901   Not tainted  (4.8.0)
> MSR: 8000000002009032 <SF,VEC,EE,ME,IR,DR,RI>  CR: 44428848  XER: 00000000
> CFAR: c0000000003a5670 SOFTE: 1
> GPR00: c0000000003a79cc c0000006283c7230 c00000000110f100 0000000000008010
> GPR04: 0000000000000000 c000000630fd7000 fffffffffffffffe 0000000000000010
> GPR08: 0000000000008010 ffffffffffffffff 000000000000000f 0000000081860c2d
> GPR12: 0000000024428244 c00000000fe00400
> NIP [c0000000003a55e4] .ext4_mb_generate_buddy+0x1a4/0x450
> LR [c0000000003a5564] .ext4_mb_generate_buddy+0x124/0x450
> Call Trace:
> [c0000006283c7230] [0000000000000002] 0x2 (unreliable)
> [c0000006283c7340] [c0000000003a79cc] .ext4_mb_init_cache+0x43c/0x830
> [c0000006283c7470] [c0000000003a8040] .ext4_mb_init_group+0x280/0x350
> [c0000006283c7540] [c0000000003a8340] .ext4_mb_good_group+0x230/0x240
> [c0000006283c75e0] [c0000000003abf8c] .ext4_mb_regular_allocator+0x2ec/0x540
> [c0000006283c7720] [c0000000003ae1c0] .ext4_mb_new_blocks+0x680/0xd80
> [c0000006283c7820] [c00000000039ee98] .ext4_ext_map_blocks+0x758/0x1d20
> [c0000006283c79d0] [c000000000365e24] .ext4_map_blocks+0x1c4/0x750
> [c0000006283c7ab0] [c000000000399230] .ext4_alloc_file_blocks.isra.6+0x170/0x3e0
> [c0000006283c7be0] [c0000000003a1e24] .ext4_fallocate+0x2f4/0xa20
> [c0000006283c7cf0] [c000000000283f54] .vfs_fallocate+0x144/0x290
> [c0000006283c7d90] [c000000000285298] .SyS_fallocate+0x58/0xd0
> [c0000006283c7e30] [c000000000009560] system_call+0x38/0x108
> Instruction dump:
> 889e0014 7f042030 5484043e 7e892378 7d2907b4 7d4900d0 7d494838 7d070034
> 7d290074 2147001f 2129003f 7f895000 <409d0008> 7d495378 7d2a07b4 2e890000
> 
> -- 
> chandan
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [BUG] softlockup observed with 64k block size on ppc64

[Chandan Rajendra]

On Tuesday, October 25, 2016 10:13:19 AM Jan Kara wrote:
> Hi!
> 
> On Mon 24-10-16 14:10:02, Chandan Rajendra wrote:
> > The following soft lockup occurs when fstests' ext4/303 test is executed with
> > 64k block size on a ppc64 machine running linux-4.8 kernel.
> 
> Thanks for report! I had a look into the code and I'm not sure why this
> happens. Is this reproducible? Did it happen also with older kernels?

Yes, This issue can be recreated every single time the test is executed. The
bug is present atleast from linux-v4.0 version. I haven't verified if the bug
is present in earlier versions of the kernel.

-- 
chandan